“Nonsense! Hot air! Balderdash!” blurted out Pavel Kroupa, an astrophysicist at the University of Bonn in Germany, as I stood at the head of the lecture hall. I was just a graduate student at the time, applying for postdoctoral research positions. I had come to Bonn to give a 45-minute talk on my investigations of the small satellite galaxies surrounding the Milky Way. I had helped develop a theory that explains why these mysterious objects are located in what appears to be a straight line stretching across the sky—an unexpected and extremely puzzling alignment. Kroupa, it appeared, was not swayed by my arguments.

Most galaxies like the Milky Way are surrounded by dozens of small satellite galaxies that orbit around them. These galaxies are extremely faint—only the brightest and closest of them have been spotted flying around the Milky Way and our next-door neighbor, the Andromeda galaxy. But these dwarf satellite galaxies do not just fly around haphazardly. Instead they all sit on a thin plane, seen edge on [see box on opposite page].

This alignment comes as a surprise. Computer simulations that model how galaxies evolve have predicted that every direction in the sky should contain roughly the same number of satellite galaxies. Such a spherical arrangement was long thought to be a natural consequence of dark matter, a mysterious substance that interacts with ordinary matter only through the force of gravity. Astronomers believe dark matter pervades the universe and plays a key role in galaxy formation and expansion of the cosmos.

Yet the puzzle of dwarf galaxy alignment has been so vexing that it has led some astronomers, including Kroupa, to question whether dark matter really exists after all. “Dark matter has failed,” he said, interrupting my talk, “since its prediction that satellites should be spherically distributed around the Milky Way is clearly in direct contradiction with what we observe.”

I was presenting a different view, one that attempts to explain the peculiar alignment of galactic satellites by pointing to cosmic structures of dark matter that are far larger than our Milky Way. Although a few skeptics like Kroupa remain unconvinced, recent work, including my own research, shows how enormous webs of dark matter can account for the unique alignment of satellite galaxies in the sky.

Missing Matter

The dark matter at the center of this debate was first postulated in an effort to explain other puzzling features of galaxies. In the 1930s the great astronomer Franz Zwicky wanted to weigh the Coma cluster, a huge group of around 1,000 galaxies. He started out by measuring the speed with which the galaxies in Coma move. To his surprise, he found enormous speeds—thousands of kilometers per second—fast enough to rip the cluster apart. Why was the cluster not tearing itself up? Zwicky concluded that the cluster must be filled with additional unseen matter that holds the galaxies together with its gravitational force. This missing substance has subsequently been named “dark matter.”

Since Zwicky's first suggestion some 80 years ago, signs of dark matter have popped up all over the universe, in nearly every galaxy observed. In our own Milky Way, astronomers infer its existence from the motion of the stars on the galaxy's outskirts. Like the galaxies in the Coma cluster, these stars move too quickly to be held in by all the matter that we see. The dozen or so dwarf galaxies of the Milky Way appear to contain even greater abundances of dark matter.

Dark matter's pervasiveness has solidified belief in its existence. In fact, most cosmologists believe dark matter constitutes around 80 percent of all matter, outweighing normal atoms by around five to one.

This abundance of dark matter implies that it should play a dramatic role in how the universe evolves. One way to study this evolution is through the use of computer models. Beginning in the 1970s, researchers in the field of computational cosmology have attempted to simulate the history of the universe using computer codes. The technique is straightforward: Define an imaginary box in a computer. Place imaginary point particles (that represent clumps of dark matter) in a near-perfect lattice inside the box. Calculate the gravitational pull on each particle from every other particle in the box and move each particle according to the net gravity it feels. Iterate this process for 13 billion years.

The strategies have grown significantly more complicated since the 1970s, but this basic technique is still used today. Four decades ago the codes could handle just a few hundred particles. Now state-of-the-art computer simulations can successfully model billions of particles in a volume approaching the size of the observable universe.

Computer simulations of the cosmos have been an incredibly useful way to investigate individual galaxies, but they have created some notable puzzles. For example, computer models conclude that the pervasive dark matter in the so-called halo that surrounds the Milky Way should pull gas and dust into individual clumps. These clumps should contract under the force of gravity, eventually forming stars and dwarf galaxies. In the case of the Milky Way, the prevalence of dark matter implies that we should expect to see thousands of small galaxies. Yet when we look out at the night sky, we observe only a few dozen. The failure to find them was first identified in the 1990s and has since become known as the missing satellites problem.

In the intervening years, astronomers have devised a few potential solutions to this dilemma. First and foremost, perhaps not all the satellites seen in simulations correspond directly to real satellite galaxies. The smallest clumps of dark matter may lack the mass (and gravitational pull) to capture gas and form stars. In this line of thinking, the observed satellite galaxies are the visible tip of a dark iceberg: hundreds, if not thousands, of dark satellites, devoid of stars, may exist in our vicinity. We just can't see them.

Second, even if small dark matter clumps do create stars, those stars may be too faint for our telescopes to see. In this scenario, as technology advances and telescopes become more sensitive, astronomers will find more satellites. Indeed, in the past seven years the number of satellites known to be orbiting the Milky Way has doubled.

In addition, the disk of the Milky Way could be blocking our view of certain satellite galaxies. This disk is essentially a dense plane of stars so bright that it looks like a continuous white fluid to the naked eye (hence, the “Milky” Way). It would be exceedingly difficult to find a satellite hidden behind the disk, just as it is difficult to see the moon during the day—the light from the disk simply drowns out the faint light from the satellite.

Taken together, these arguments largely settled the missing satellite problem for most astrophysicists and saved the idea of dark matter from one of its most serious observational challenges. Yet the peculiar alignment of satellite galaxies continued to baffle researchers.

Return of the Dwarf Threat

In several papers in the late 1970s and early 1980s, Donald Lynden-Bell, an astrophysicist at the University of Cambridge, noted that many of the satellite galaxies orbiting the Milky Way appeared to sit on a single plane. How could this odd arrangement be explained? In 2005 Kroupa and his group at Bonn convinced the world that the alignment could not be random. They assumed that dark matter satellites were evenly distributed around the Milky Way, as the computer simulations predicted, and that only one in 100 of these dwarfs was large enough to create stars and visible galaxies. Given these entirely reasonable assumptions, they asked, how often would we expect to find a system like the Milky Way, where the illuminated galaxies happen to be located all in a row? The answer caused an earthquake in cosmology: the probability was less than one in a million.

If dark matter guided the formation of galaxies, Kroupa argued, the dwarf satellites would never all be found on this one impossible plane. In the paper describing his results, Kroupa put forth his own solution. The only way out, he wrote, was if the Milky Way's satellites did not form as a consequence of the clumping of dark matter. Dark matter, he said, does not exist.

As a good theorist, Kroupa proposed an alternative. He suggested that satellites were galactic debris, the remains of an older progenitor galaxy that long ago flung past the Milky Way. Just as an asteroid breaks up and leaves a trail of debris as it flies through Earth's atmosphere, perhaps the satellites of the Milky Way similarly had their origins in material stripped from a larger progenitor.

For example, Kroupa said, when we look out into the cosmos, we can see that a number of colliding galaxies show long bridges of material known as tidal arms. Often the tidal arms contain small dwarf galaxies that condense out of the streaming material. Under the right conditions, the nature of the ripping ensures that the stripped material will end up in a thin plane, just like the satellites of the Milky Way.

Kroupa's explanation was elegant, simple—and above all, controversial. It quickly came under attack. For one, the stars in the satellites of the Milky Way are moving far too quickly to be held together by ordinary matter alone. Dark matter must be holding them together, just as it holds the Milky Way together. (In fact, observations suggest that the dwarf satellites of the Milky Way are among the most dark matter–dominated galaxies in the universe.) The tidal dwarf galaxy scenario implies that these galaxies are devoid of dark matter, leaving open the question of what keeps them from flying apart.

Secondly, just as car crashes destroy cars, collisions between disk galaxies destroy the disks. The final result of a galactic collision is almost always a formless blob of stars. The Milky Way has a crisp structure and a fairly thin disk. We observe no indication that it suffered through any merger or collision in the recent past.

The Dark Web

An alternative solution to the unusual alignment of dwarf galaxies requires looking farther out into the cosmos. The computational simulations that began in the 1970s do not just model the evolution of individual galaxies. They model huge volumes of the universe. When we explore these simulations on the largest scales, we see that galaxies are not randomly distributed. Instead they tend to aggregate into a well-defined filamentary network known as the cosmic web. We clearly see the predicted structure when we look up to the skies with large-scale astronomical surveys.

The cosmic web is composed of magnificent sheets of millions of galaxies, hundreds of millions of light-years across. Cigar-shaped filaments connect these sheets. In between the filaments lie massive voids where no galaxies reside. Large galaxies such as the Milky Way tend to anchor the web at spots where multiple filaments intersect [see box on opposite page].

As a graduate student at Durham University in England, I had been creating computer simulations of these dense regions when I brought a plot of recent results into the office of my research adviser, Carlos Frenk. The model I had been working on traced the formation of the Milky Way and its environs for the past 13 billion years of cosmic history. Frenk scrutinized the plots for a moment, shook the papers and exclaimed, “Drop everything! The satellite galaxies you are studying are all sitting on Kroupa's impossible plane!” Our model was not reproducing the earlier predictions of computer simulations—an evenly distributed halo of satellite galaxies around the Milky Way. Instead the computer was predicting the formation of a plane of satellites that was very close to what astronomers observe. We felt that our simulations were beginning to crack the mystery of how the dwarf satellites came to adopt such an odd configuration.

“Why don't you trace the satellites back in time and see where they came from?” Frenk proposed. We had the final result; now it was time to examine the intermediate steps in the simulation.

When we examined the simulation in reverse, we saw that the dwarf satellites did not originate in the region immediately surrounding the Milky Way. They tended to come together a little farther away, inside of filaments in the cosmic web. Filaments are regions of the cosmos with higher densities than the cosmic voids; as such, they will attract nearby dust and gas and collect them into nascent galaxies.

Once these dwarf galaxies form, gravity pulls them in the direction of the most massive nearby region—in our case, the Milky Way. Because the Milky Way lies at a node where filaments intersect, the dwarf galaxies travel through the filament that birthed them as they accelerate in our direction. Filaments, in other words, serve as cosmic superhighways of dark matter. When we gaze up at the sky and see dwarf galaxies in a single plane moving in the same direction, we are essentially looking at oncoming galactic traffic.

A New Test

Some scientists such as Kroupa remain skeptical. Computer models seem to reproduce the observed conditions around the Milky Way with sufficient precision, but the general theory should be also able to describe the neighborhood around other galaxies.

The theory faces a new test. In January 2013 astronomers mapping the regions around the nearby Andromeda galaxy found an even thinner sheet of satellites: a vast plane one million light-years across and just 40,000 light-years thick—around the same dimensions of a laptop computer. The sheet also appears to be rotating in just the way that Kroupa's tidal scenario would predict. Computer simulations such as my own, however, have not yet been able to reproduce the alignment of galaxies that we see around Andromeda.

Yet the serious problems with Kroupa's tidal theory remain—it, too, is at odds with observations. History has shown that in stalemates such as these, definitive solutions will only come with more data. As Albert Einstein once remarked, “Nature did not deem it her business to make the discovery of her laws easy for us.”